Fig. 4: Illustration of the TRAPT framework by using downregulated genes from ESR1 knockout experiments in cases of gastric cancer and MCF7 breast cancer. | Nature Communications

Fig. 4: Illustration of the TRAPT framework by using downregulated genes from ESR1 knockout experiments in cases of gastric cancer and MCF7 breast cancer.

From: TRAPT: a multi-stage fused deep learning framework for predicting transcriptional regulators based on large-scale epigenomic data

Fig. 4

a Scatter plot displaying values of the average normalized activities of 1358 TRs for the upregulated and downregulated gene sets. The size of each data point represents the magnitude of the average normalized activity, while the colors represent different categories of TRs: TFs (blue), TcoFs (green), and CRs (yellow). b This network diagram was derived from predictions of protein–protein interactions in the STRING database. The size of nodes represents their degrees, and the thickness of edges represents the probability of interaction. c This heat map was derived from results of the co-expression analysis of TCGA breast cancer, with the depth of the colors indicating the degree of correlation. d Bar chart showing the normalized activity scores of the TRs. ESR1, GATA3, FOXA1, and EP300 were among the top 10 TRs, while HDGF was ranked last. ESR1 had the highest score. e Comparison of D-RP scores between the queried (n = 494) and background genes (n = 25646), revealing significant differences for all TRs except HDGF. Middle line inside each box represents the median, upper and lower bounds of the box represent the third and first quartiles, respectively. P-values are calculated by the two-sided Mann-Whitney U test without adjustments. f Aggregated profiles of enhancer marks. Except for HDGF, the marks of all TRs near the queried gene were significantly higher than those near the background gene. P-values are calculated by the two-sided Kolmogorov-Smirnov test without adjustments. g Heat maps of the activity matrix before and after integrating REA scores, demonstrating the differentiation between the queried and background gene sets. We randomly selected 10,000 genes for visualization. h Genome browser displaying the tracks of ESR1, GATA3, FOXA1, EP300, and HDGF near the genes ESR1, GREB1, TFF1, and CCND1. We selected the tracks of the 10 epigenomic samples with the largest weights in the reconstructed network for ESR1, shown as ATAC (blue) and H3K27ac (green) tracks.

Back to article page